NEW: quick chart & thread to clear up some confusion following @Peston’s digging into ONS modelled estimates:
It’s true that ONS modelled incidence series has issues, and the revisions can provoke doubts.
BUT that doesn’t mean gov was flying blind when it announced 2nd lockdown
It’s true that ONS modelled incidence series has issues, and the revisions can provoke doubts.
BUT that doesn’t mean gov was flying blind when it announced 2nd lockdown
https://twitter.com/Peston/status/1336246651799937024
Let’s set aside that ONS series, and instead look at:
• Imperial REACT study
• Covid symptom tracker
• ONS’s model-free weighted estimates (not subject to revisions)
• The dashboard
Each as it stood when lockdown was announced.
All pointed to rapidly increasing infections
• Imperial REACT study
• Covid symptom tracker
• ONS’s model-free weighted estimates (not subject to revisions)
• The dashboard
Each as it stood when lockdown was announced.
All pointed to rapidly increasing infections
tl;dr
The way ONS models incidence can a) give a misleading shape to recent days of the outbreak, and b) cause confusion where estimated prevalence is retrospectively altered.
But there’s no doubt the virus was spreading rapidly in late October when lockdown was announced
The way ONS models incidence can a) give a misleading shape to recent days of the outbreak, and b) cause confusion where estimated prevalence is retrospectively altered.
But there’s no doubt the virus was spreading rapidly in late October when lockdown was announced
Further reading:
• @ChrisGiles_ was highlighting issues with the modelled ONS series in October
• @danielhowdon has many threads on the subject in recent weeks
• @SarahDRasmussen has also gone into detail
• @ChrisGiles_ was highlighting issues with the modelled ONS series in October
https://twitter.com/ChrisGiles_/status/1319620517796741120
• @danielhowdon has many threads on the subject in recent weeks
https://twitter.com/danielhowdon/status/1334854042825191426
• @SarahDRasmussen has also gone into detail
https://twitter.com/SarahDRasmussen/status/1336190871306465281
And a final note: the @ONS has been brilliant throughout the pandemic, providing more high quality, timely and robust data than any other national statistical body I’ve encountered worldwide.
The modelling issue here is a slip-up, but shouldn’t tarnish the rest.
The modelling issue here is a slip-up, but shouldn’t tarnish the rest.
Final note:
None of the data shown in my chart has been revised down. ONS runs a model *using the data shown in the red line* which does weird things including overwriting old data with new data every week, but the *actual prevalence data they measure* (red line) has not changed
None of the data shown in my chart has been revised down. ONS runs a model *using the data shown in the red line* which does weird things including overwriting old data with new data every week, but the *actual prevalence data they measure* (red line) has not changed
• • •
Missing some Tweet in this thread? You can try to
force a refresh
